1 / 24

INTEGRATION AND CONVERGENCE IN EUROPEAN ELECTRICITY MARKETS

INTEGRATION AND CONVERGENCE IN EUROPEAN ELECTRICITY MARKETS C. A. BOLLINO - D. CIFERRI - P. POLINORI Department of Economics, Finance and Statistics - University of Perugia 30 th USAEE/IAEE NORTH AMERICAN CONFERENCE Oct 9-12, 2011,WASHINGTON, DC. Research Questions.

anthea
Download Presentation

INTEGRATION AND CONVERGENCE IN EUROPEAN ELECTRICITY MARKETS

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. INTEGRATION AND CONVERGENCE IN EUROPEAN ELECTRICITY MARKETS C. A. BOLLINO - D. CIFERRI - P. POLINORI Department of Economics, Finance and Statistics - University of Perugia 30th USAEE/IAEE NORTH AMERICAN CONFERENCE Oct 9-12, 2011,WASHINGTON, DC

  2. Research Questions • The liberalization process of electricity markets in Europe is more than a decade old. • In the meantime, mergers and restructuring of big players in generations at the international level make likely that decisions and price strategies are taken simultaneously on several markets, based on a common set of available information. • There is large evidence that organized electricity spot markets are far from the ideal competitive model and therefore all sorts of behaviors and shocks may influence price formation in those markets, ranging from international fuel price, local meteo, and local market power behavioral shocks. • In this respect, we want to investigate whether there exists some information signaling among different European markets.

  3. Data • Data of (the log of) Electricity prices for DE, AT, FR, and IT are used • Estimation sample period: from 1/4/2004 to 8/3/2010

  4. The Empirical Strategy According to stochastic definitions of convergence and common trends based on cointegration analysis (Bernard, 1991), a necessary (but not sufficient) condition for convergence among countries and/or markets is that there should be n-1 cointegrating vectors for a sample of n countries or markets. We use a multivariate specification for the four equation of electricity spot prices according to a Vector Auto Regressive (VAR) process of order p where

  5. The Empirical Strategy The previous equation can be represented in its isomorphic Vector Error Correction (VEC) form: On the basis of the rank of matrix , it is possible to identify different long-run equilibrium path for the electricity prices in the models. In any intermediate result with a reduced rank of matrix we have a long run representation of the integration process between markets.

  6. The Empirical Strategy In the presence of cointegration, the rank of matrix is reduced to r < 4 and can be decomposed as: ∏y = αβ’ where: αshows the feedback coefficients β shows the theory based long-run relationship coefficients

  7. Preliminary Analysis and VECM specification We test for unit root behavior of each of the four series (ADF and PP). In each case, we are unable to reject the unit root-null hypothesis at conventional nominal significance levels. VECM specification: we allow for a lag length of 22 (in levels) and an unrestricted constant term in the VECM specification. The choice of optimal lag length has been done according with the lag-exclusion (the initial specification includes 31 lags) method at 10% level of significance The results of the main univariate and multivariate diagnostic tests (for serial correlation, normality, heteroskedasticity and ARCH components) indicate that estimated residuals resemble white-noise process in a satisfactory way at both single equation and system level.

  8. Long- run properties

  9. Hypothesis testing on Beta vectors LR χ2(3)=5.925[0.115]

  10. The long run equilibrium conditions • This representation implies three different bilateral integration processes between Germany market (the common trend) and the other ones. • Using a standard -distributed LR ratio test with 3 degrees of freedom, the test statistics (5.925), calculated using the Bartlett small-sample correction (with estimated factor of 4.72), indicate that the restrictions are not rejected by the data at the usual significance levels (p-value of 0.115)

  11. Granger-Causality Test statistic l = 1.6592; pval-F( l; 57, 8264) = 0.0014 Test statistic: c = 828.0805 pval-Chi( c; 3) = 0.0000 We test also for Granger-Causality in the whole system in order to verify if DE-prices “Granger-cause” the other variables in the model  Thus, we perform the usual F-test on the significance of lagged values of DE in the equations of AU, FR and IT. Under H0: " DE" does not Granger-cause « AU, FR, IT" The test confirms that DE cause “AU, FR, IT" As robustness, we develop also a test for Instantaneous Causality. H0: No instantaneous causality between “DE" and “AU, FR, IT“ USAEE/IAEE North American Conference, Washington, DC - October 9-12, 2011

  12. Cointegration vectors

  13. Persistence Profiles • Persistence profiles (Pesaran and Shin,1996) consider the effect of a system-wide shocks on cointegrating relations, and it is given by: • The value of this profile is equal to 1 on impact, but should tend to zero as N→ ∞ if βj is indeed a cointegrating vector. • The persistence profiles are also useful in the case of time series that are close to being I(1). If this is the case, the persistence profiles eventually converge to zero, but can be substantially different from zero for a protracted period. • The persistence profiles, viewed as a function of N, provides important information on the speed at which the system-wide shocks effect disappears.

  14. Persistence Profiles Simulation period: 5 months

  15. The restricted VECM • The next step is to estimate a parsimonious VECM by 3SLS in which statistically irrelevant parameters are deleted through the SER/TP method. The AIC criterion with t=1,6 is used as a significance threshold level for short-run parameters. • The choice is motivated by the opinion that, in the reduction process of the model, it is preferable to maintain the coefficients with uncertain significance.

  16. VECMmodel estimated 3SLS- The long run Equations are obviously affected by the cointegration residuals which identify the long run convergence equilibrium between each market and the German system. Moreover the (absolute) values of the feedback coefficients indicate that the speed of adjustment towards equilibrium is higher for Austria. Finally, there is evidence of some influence of each error on the other market equations. ,

  17. The Long Run

  18. Dynamic Simulation • If we consider the VECM representation, we can always rewrite the model in MA representation as: • With an appropriate choice of the matrix C we are able to perform FEVD analysis separating (n-r) permanent shocks from (r) transitory shocks

  19. Dynamic Simulation: FEVD Simulation period: 5 months

  20. Conclusions • German market behavior appears as the common trend for other regional markets, thus providing signaling information. • This can be explained in two ways: (i) DE is the largest market in Central Europe and it is taken as a reference; (ii) pricing in electricity markets is dominated by peak-load plants, which typically exhibit CCGT technology (i.e. gas fired) and gas marginal price is largely influenced by German market operators • Persistence appears to be higher in FR . This is no surprise, given that the French electric system is the most un-flexible (because of its very high nuclear share).

  21. Conclusions • The permanent shock is associated to the common trend of the system (that is the German electricity spot price) and represents the global-external shocks that hit in a symmetric way the other markets. • Individual temporary shocks identify idiosyncratic disturbances. Idiosyncratic shocks are then aggregated so as to quantify the overall relevance of regional factors in explaining spot prices fluctuations. • The figures represent the percentage of the variance of each variable of the system explained by global, regional and idiosyncratic shocks, where the latter (in italics) are expressed as a percentage of regional disturbances. • The last column (mean) presents the average contribution of the shocks over the entire simulation period (5 months).

  22. Conclusions FEVD analysis shows that IT is the market with lowest share of global shock compared to other countries. Thus the signaling effect of global shocks in price formation is the least important in the Italian case The fact that roughly 1/4 of FEVD is not explained by a global shock (which is typically the fuel price shock) indicates that there are other factors, like non competitive strategic behavior, influencing equilibrium prices, which motivates future research.

More Related